GPU Boost 2.0 - Temperature Target

We now want to explore what changing the temperature target does to performance and fan noise.



First, we increased the temperature target by +10°C, to 90°C. This resulted in a performance increase of 2%, a relatively small improvement for much higher fan noise (going from 38 dBA to 51 dBA).



When set to the maximum possible temperature target of 94°C, an additional 0.5% performance is gained (+2.4% over the baseline at 80°C). This is certainly not worth the added noise (55 dBA now), at least for most users.



You may also reduce the top temperature target to optimize the acoustic experience. The card loses a lot of performance with a 70°C target (-26%), but runs significantly quieter (30 dBA). If you value low noise, this option might still be viable as the card should still be fast enough to run most games.

Please also observe how on all these tests, the performance difference varies greatly between the games we tested.



In a final test, we set the maximum temperature target to 94°C while also setting the maximum power target to 106%. The card can now draw as much power as allowed by NVIDIA while also running as hot as allowed by NVIDIA.

As a result, we see a performance increase of 2.33% over the out-of-the-box baseline. Please note that this is 0.1% lower than when we set 94°C as the target without a power increase. It is absolutely possible that the difference is due to random variations in benchmark runs. However, another possible explanation is that the increased power limit makes the card draw more power, which results in higher temperature, which means the card runs into the temperature limit a tiny bit sooner, so the card will drop clocks again.